Risk stratification, genomic data and the law

Abstract

Risk prediction models have a key role in stratified disease prevention, and the incorporation of genomic data into these models promises more effective personalisation. Although the clinical utility of incorporating genomic data into risk prediction tools is increasingly compelling, at least for some applications and disease types, the legal and regulatory implications have not been examined and have been overshadowed by discussions about clinical and scientific utility and feasibility. We held a workshop to explore relevant legal and regulatory perspectives from four EU Member States: France, Germany, the Netherlands and the UK. While we found no absolute prohibition on the use of such data in those tools, there are considerable challenges. Currently, these are modest and result from genomic data being classified as sensitive data under existing Data Protection regulation. However, these challenges will increase in the future following the implementation of EU Regulations on data protection which take effect in 2018, and reforms to the governance of the manufacture, development and use of in vitro diagnostic devices to be implemented in 2022. Collectively these will increase the regulatory burden placed on these products as risk stratification tools will be brought within the scope of these new Regulations. The failure to respond to the challenges posed by the use of genomic data in disease risk stratification tools could therefore prove costly to those developing and using such tools.

Of Screening, Stratification, and Scores

“…The estimate provided by PRSs is calculated based on the individual’s genotype profile in comparison with the relevant GWAS data [ 82 ]. The ever-growing use and availability of large quantity of genomic and health-related data from which the foregoing can be discovered and furthered, the identified advantages of preventive medicine, and the increasing personalization of medicine have emphasized the use and potential usefulness of polygenic risk scores (PRS) in risk stratification and screening practices and programmes [ 85 ].…”

“…However, their clinical utility and analytical validity are still to be improved and refined for wider clinical use in risk prediction, treatment response, and prognosis [ 83 ]. These scientific and technical limitations complicate their current definition within the regulatory frameworks, as their use for medical purposes is uncertain under the law [ 12 , 85 ]. This uncertainty extends to all jurisdictions.…”

“…While in some jurisdictions, PRSs could be deemed to fall under MD regulatory oversight, in others, they would not [ 88 , 89 , 90 , 91 , 92 ]. Moreover, as MDs, they could be considered standalone MDs or accessories, or even in vitro (IV) MDs [ 85 , 93 ].…”

Genetic database software as medical devices

“…The EU IVDR, for example, adopts a deterministic approach to classify the risk of genetic IVDs, assigning all genetic tests to the same medium–high risk class regardless of whether the test is predictive or diagnostic (EU IVDR, Annex VIII). This classification mandates “comprehensive requirements for clinical evidence of performance and effectiveness” before a genetic IVD can receive a CE Marking and be placed on the market (Hall et al., 2018). This may result in procrustean regulation of genomic software: strict rules if they qualify as IVDs or no rules at all.…”

“…To conclude, we recommend strategies for ensuring and demonstrating that publicly available precision medicine software applications are safe and effective when used to support the predictive testing, diagnosis, and treatment of patients. While we focus on software developed to enhance the interactivity of variant databases in support of precision medicine, our discussion will also be relevant to other genomic software tools, including clinical decision support intended to provide users “with knowledge and person‐specific information, intelligently filtered or presented at appropriate times, to enhance health and health care” (Senger & O'Leary, 2018), risk prediction algorithms (Hall et al., 2018), and in silico prediction tools (Prawira, Pugh, Stockley, & Siu, 2017). Our discussion is limited to the quality and regulation of genomic software .…”